A Case Reasoning Method Based on Dynamic Knowledge Representation Learning

Abstract

A case reasoning method based on dynamic knowledge representation learning, including the following steps: 1) Obtain all relevant data of cracked cases, divide them into five types of entities: personnel, cases (events), objects, locations, and institutions, and extract corresponding relationships ; 2) Store the extracted events in the form of time, entity, and relationship as a quadruple format, and store them in the graph database; 3) Optimize the hyperparameters of Gaussian process regression based on the gradient descent algorithm; 4) Use It is a recurrent neural network model, which performs cyclic event reasoning on quadruple data; 5) uses graph databases to perform first-degree and second-degree relationship search, and performs link prediction based on the search results; 6) evaluates the prediction results and Sort. The present invention uses a dynamic knowledge representation learning algorithm to embed entities and relationships in quadruples, conducts training and learning on the basis of the constructed knowledge map, deduces possible criminal suspects, and simplifies police work.

Description

technical field [0001] The invention relates to knowledge graphs, police big data, and deep learning, and in particular to a case reasoning method based on dynamic knowledge representation learning. Background technique [0002] With the continuous improvement of the city's informatization level and the rapid development of science and technology, the popularity of artificial intelligence in today's society is getting higher and higher, and its influence is becoming more and more far-reaching. It plays a pivotal role and greatly facilitates people's daily work and life. [0003] At the Yunqi Conference in October 2016, Hangzhou Municipal Bureau of Economy and Information Technology, Alibaba and other enterprises launched the Lihangzhou "Urban Data Brain" project, preparing to rely on big data, cloud computing and artificial intelligence to integrate government data and public data. , enterprise data, and Internet data, use informatization and intelligent means to build a ci...

AI Technical Summary

Problems solved by technology

[0004] However, limited by the many deficiencies of the current deep learning technology, it can only solve tasks with regular data and relatively simple tasks. In some situations with complex structures, such as police research and judgment, it is difficult for deep learning technology to play a role, and it is still necessary to rely on Judging by human experience

Images